An operations manager of a computer system monitors a fault that influences a service operated on the system, and abnormality pre-indicating a portent thereof. Further, when the fault or the abnormality is detected, the operations manager analyzes causes thereof, and takes measures when necessary.
Software for assisting such a service management work includes monitoring software. The monitoring software includes software for assisting fault detection and a cause analysis. First of all, the monitoring software has a function for communicating with hardware or software of a monitoring target, and collecting log data showing a system operation situation. The log data includes numeric data such as a CPU utilization rate, or character string data such as a text message presenting an application state. In recent years, monitoring software for collecting a variety of the log data to allow search is also present. Second of all, the monitoring software has a function for processing the log data into the form of a line graph or the like, and displaying the processed form on a screen of an operations manager terminal.
A monitoring mechanism has been so far constructed for every log data or every service. However, in connection with an influence of system integration by recent virtualization technology, an integrated monitoring mechanism for unitarily managing a plurality of services has been required.
One embodiment of monitoring software that achieves such integrated monitoring includes a monitoring platform. In the monitoring platform, a function is provided for putting different kinds of log data of different services into one place and managing the log data. In the monitoring platform, upon request from the operations manager, the Log data stored on a log database are searched, and the results are presented to the operations manager.
In the log database, capability of storing different formats of data is required depending on kinds of services or log data. Technology for flexibly storing such data exists. Patent document 1 (JP-A-2011-134342) discloses a method for, targeting a database preliminarily needing a definition of a schema (stored data structure), such as a relational database (hereinafter, RDB), storing schema data variable to a defined fixed schema. In the technology, first, a plurality of data field columns for variable data are preliminarily prepared upon defining the fixed schema. Then, definition information in different data formats for every data type is managed outside the schema, thereby allowing storage of values in different data formats to an identical field column. As a method different from the method in Patent document 1, a method also exists for storing in a hash table a key and a value of data in different formats to store a binary of table data thereof in a single column of a relational database. Furthermore, a new database provided with schema flexibility as referred to as a document-oriented database has recently appeared. The document-oriented database provides a basic function almost similar to the function of RDB, such as storage, search and indexing. However, in the document-oriented database, data are not stored in a tabular format as in the RDB, but are stored in a combination of a key and a value. Therefore, a preliminary schema definition becomes unnecessary. In the document-oriented database, the method for storing the key and the value, as is different from the art disclosed in Patent document 1 for preliminarily defining what data to be entered into what column, has an advantage of capability of corresponding also to a case where a data column definition or sequence is changed later.
Meanwhile, as technology for reducing human resources in configuration, a method exists for automating a general database configuration. Patent document 2(JP-A-2003-228570) and Patent document 3(JP-A-H6-215037) disclose arts for evaluating a database search index from search frequency to an attribute to perform automatic adjustment. Patent document 4(JP-A-H7-85093) discloses an art for evaluating frequency of utilization frequency of a preregistered index candidate from search frequency to sequentially select an index from a superordinate. Patent document 5(JP-A-H10-111819) discloses an art for automatically generating an index in which a cumulative value of search response time and insertion response time is minimized.